How I Built Algocdk: An Algorithmic Trading Platform for Deriv Traders

Most retail traders I know do the same thing every day: stare at a chart, wait for a setup, click buy, click sell, repeat. It works until you are tired, distracted, or asleep, and the markets never are.

So I built Algocdk, an algorithmic trading platform for Deriv traders. It lets you analyze markets, build indicators, run bots 24/7 and trade, all from one place, with no installs.

You can try it here: algocdk.site

In this post I'll walk through why I built it, what it does, and how you can build your own custom indicator with it.

Why I built it

Deriv offers a great API and a huge range of markets: Forex, crypto, commodities, indices and its own synthetic indices that run around the clock. But the tooling around it was split across many places. Charts lived in one tab, analysis in a spreadsheet, bots in another tool, and nothing talked to anything else.

I wanted a single developer-friendly terminal where I could:

• see live Deriv market data
• analyze it with real tools
• write my own indicators in JavaScript
• turn an idea into a running bot without babysitting it

That became Algocdk.

The architecture in a nutshell

At a high level, the platform has three layers:

1. Market data layer. A live connection to Deriv APIs streams ticks and candles into the app in real time.
2. Analysis and charting layer. Interactive charts, indicators and analysis tools sit on top of that stream.
3. Automation layer. Bots run inside isolated containers so each strategy executes on its own, continuously watching the market.

The isolation part was a deliberate decision. Every bot runs in its own container, so a bug or a bad strategy in one bot can't take down the others or the rest of your account setup.

Features

Here is everything Algocdk does today.

1. Live markets and charting

• Real-time quotes across Forex, Crypto, Synthetic Indices, Commodities and Global Indices
• A full candlestick chart with candle, line and area views
• Timeframe switching and an indicators menu, plus a compare mode
• A market overview with crypto, forex, indices and commodities (Bitcoin, Ethereum, EUR/USD, S&P 500, Gold, Crude Oil and more)
• Synthetic Indices such as Volatility 10, 25, 50, 75 and 100, and Boom/Crash, streaming a tick every second, 24/7

2. A trading platform for Deriv contract types

You can trade directly from the platform across the main Deriv contract types:

• Rise / Fall
• Higher / Lower (barriers)
• Multipliers
• Accumulators
• Digital Options
• Digits trading

3. Market analysis tools

This is the part I'm most excited about, because it turns gut feeling into data:

• Rise/Fall market analysis
• Higher/Lower market analysis
• Digit Analysis for studying last-digit distributions
• Digit Lab for experimenting with digit strategies
• Digit Hit Map for visualizing which digits are hitting
• Market News and an Economic Calendar so you know when volatility is coming

4. Automated trading bots

Algocdk ships with strategy types that cover the common approaches:

Strategy	What it does
Momentum	Trend-following with real-time signal validation
Breakout	Catches volatility expansions with adaptive risk management
Mean Reversion	Exploits deviations from historical averages
Grid	Places buy and sell orders at set intervals in range-bound markets
Scalping	Targets quick, small moves
Build Your Own	Visual bot builder, or code your own in JavaScript or Python

On the bot side you also get:

• A Bot Marketplace (Bot Store) to browse and deploy ready-made bots
• DBot support
• An AI Over/Under Bot
• Copy Trading
• Isolated bot containers, one per bot, running continuously
• Walk-forward backtesting, so you can validate a strategy on real data before risking real money

5. Built for developers

I built this as a developer first, so there is a lot of room to extend it:

• Custom JavaScript indicators, with an Indicator Template to start from
• A Developer Guide documenting how to build on the platform
• A Terminal CLI for people who live in the command line
• A Trading Tools section
• Programmable access built on Deriv APIs

6. AI assistance

• An AI Assistant to help you reason about markets and strategies
• A Kiro Bot Assistant to help you build bots

7. Community

• A community hub for sharing trade ideas and technical analysis on forex, crypto, futures and more
• Explore Sites, where users can browse public sites
• A support center and a feedback page, because I read all of it

8. Getting started takes about three steps

1. Create an account. Start on a zero-risk demo account and explore everything.
2. Connect your Deriv API token. Set your own risk limits.
3. Deploy a bot. Pick a strategy or build your own, and let it run.

It works in the browser and is set up as an installable mobile web app, so it feels at home on a phone too.

Build your own indicator (using NEXUS AI as the example)

An Algocdk indicator is one JavaScript object that computes values from candles and then draws them on the chart. To show how it works, I'll use NEXUS AI, an indicator I wrote that detects chart patterns, candlestick patterns, support and resistance zones, smart money concepts and overall market behavior. It's about 1,000 lines, but the structure is small enough to learn in a few minutes.

The example follows the calculate → draw indicator API. Check the Developer Guide and the Indicator Template for the exact loader details.

The shape of an indicator

The whole file is a single object literal wrapped in parentheses, so it evaluates to an object you can paste into the indicator loader:

({
  name: "NEXUS AI — Pattern & Behavior Engine",
  window: 1,          // 1 = draw on top of the main chart
  color: "#00f5d4",
  defaultParams: { sensitivity: 2, rsiPeriod: 14, lookback: 60 /* ... */ },

  calculate: function (data, params) { /* runs on ALL candles */ },
  draw: function (ctx, values, pad, spacing, yFunc, params, visibleStartIdx) { /* paints */ },
})

Field	What it does	In NEXUS AI
name	Label shown in the indicator list	"NEXUS AI — Pattern & Behavior Engine"
window	Where it renders; 1 overlays the main chart	1
color	Default accent color	#00f5d4
defaultParams	User-editable settings and their defaults	sensitivity, RSI/ATR periods, show/hide toggles for each layer
calculate(data, params)	Receives every candle and does the math; returns one value per candle	Runs all detectors, caches results, returns EMA 20
draw(ctx, values, pad, spacing, yFunc, params, visibleStartIdx)	Paints onto a canvas for the visible candles	Draws EMAs, zones, patterns, labels and the info panel

How the two functions work together

1. calculate receives data, an array of candles, each with open, high, low and close.
2. It computes everything once and returns an array the same length as data, using null where a value doesn't exist yet (for example, during an EMA's warm-up). NEXUS returns data.map(() => null) if there are fewer than 30 candles.
3. It stores richer results on the object itself (this._computed) so draw can read them.
4. draw receives a canvas 2D ctx and a few helpers: pad (left padding in pixels), spacing (pixels per candle), yFunc (converts a price to a y coordinate) and visibleStartIdx (which candle the screen starts at).
5. It converts candle indexes to x positions, then draws with normal canvas calls like moveTo, lineTo, fillRect and fillText.

Your first indicator: an EMA cross

Here is a small indicator that uses the same structure as NEXUS AI. It draws a fast and a slow EMA and marks the candles where they cross. Paste it into the indicator loader to try it.

({
  name: "EMA Cross",
  window: 1,
  color: "#00f5d4",
  defaultParams: { fast: 9, slow: 21 },

  calculate: function (data, params) {
    if (!data || data.length < params.slow) return data.map(() => null);

    // Same EMA idea NEXUS uses in _indicators.ema
    const ema = (period) => {
      const k = 2 / (period + 1);
      let prev = data[0].close;
      return data.map((c, i) => {
        prev = i === 0 ? c.close : c.close * k + prev * (1 - k);
        return prev;
      });
    };

    const fast = ema(params.fast);
    const slow = ema(params.slow);

    // Find crossovers once, here, not on every redraw
    const crosses = [];
    for (let i = 1; i < data.length; i++) {
      if (fast[i - 1] <= slow[i - 1] && fast[i] > slow[i]) crosses.push({ idx: i, bias: "bullish" });
      if (fast[i - 1] >= slow[i - 1] && fast[i] < slow[i]) crosses.push({ idx: i, bias: "bearish" });
    }

    this._computed = { fast, slow, crosses, data };  // cache for draw()
    return fast;                                      // one value per candle
  },

  draw: function (ctx, values, pad, spacing, yFunc, params, visibleStartIdx) {
    if (!this._computed) return;
    const { fast, slow, crosses, data } = this._computed;
    const start = visibleStartIdx || 0;
    const end = start + values.length;
    const x = (i) => pad + (i - start) * spacing + spacing / 2; // candle index -> pixels

    const line = (arr, color) => {
      ctx.strokeStyle = color;
      ctx.lineWidth = 1.5;
      ctx.beginPath();
      let started = false;
      for (let i = start; i < end; i++) {
        if (arr[i] == null) continue;
        if (!started) { ctx.moveTo(x(i), yFunc(arr[i])); started = true; }
        else ctx.lineTo(x(i), yFunc(arr[i]));
      }
      ctx.stroke();
    };

    ctx.save();
    line(fast, "#00f5d4");
    line(slow, "#ffa500");

    crosses.forEach((c) => {
      if (c.idx < start || c.idx >= end) return;   // skip off-screen signals
      const bull = c.bias === "bullish";
      ctx.fillStyle = bull ? "#00e676" : "#ff5252";
      ctx.font = "12px sans-serif";
      const y = bull ? yFunc(data[c.idx].low) + 14 : yFunc(data[c.idx].high) - 6;
      ctx.fillText(bull ? "▲" : "▼", x(c.idx) - 5, y);
    });
    ctx.restore();
  },
})

The parts map straight onto NEXUS AI:

• defaultParams holds the fast and slow periods, like rsiPeriod and atrPeriod in NEXUS.
• calculate does all the math and stores it in this._computed.
• draw only paints, using a start offset so signals line up with the visible candles.
• Crossovers are plain objects (idx, bias), the same way NEXUS stores { idx, type, bias, strength } for each candlestick pattern.

Change the periods, swap EMA for another average, or add a volume filter, and you already have your own indicator.

What NEXUS AI teaches: patterns worth copying

Once the basics click, NEXUS AI shows how to scale an indicator without turning it into a mess.

1. Split the work into small detectors. Each concern lives in its own sub-object (_candlePatterns, _chartPatterns, _srZones, _smc, _behavior) with a detect() or analyze() method. calculate just calls them in turn.
2. Keep helpers on the object. _utils (averages, pivots, slope, similarity checks) and _indicators (RSI, EMA, ATR, MACD, ADX) are reused by every detector. The underscore marks them as internal.
3. Detectors return plain data, not drawings. A detector produces objects like { idx, type: "Bullish Engulfing", bias: "bullish", strength: 3 }. Only draw decides how they look.
4. Compute in calculate, render in draw. draw can run every time the user pans or zooms, so the heavy work happens once and is cached in this._computed.
5. Convert global indexes to screen positions. NEXUS uses ix(globalIdx) = pad + (globalIdx - startIdx) * spacing + spacing / 2 and an isVisible() check, so it only draws what's on screen.
6. Give every layer a toggle. Params like showSRZones, showSMC and showCandlePatterns are tested with params.x !== false, so each layer can be switched off without breaking the rest.
7. Draw in layers and wrap it in ctx.save() / ctx.restore(). NEXUS paints EMAs first, then zones, smart money concepts, chart patterns, candle patterns and finally the info panel, so important labels end up on top.

Here is a trimmed detector from the file to show the style. It flags a bullish engulfing candle:

if (
  u.isBearish(p) && u.isBullish(c) &&
  c.open < p.close && c.close > p.open &&
  u.bodySize(c) > u.bodySize(p)
) {
  patterns.push({ idx: i, type: "Bullish Engulfing", bias: "bullish", strength: 3 });
}

Every pattern in NEXUS follows that same recipe: a condition on the last few candles, then a record with an index, a name, a bias and a strength.

What NEXUS AI detects, and where to extend it

• Candlestick patterns: Doji, Hammer, Shooting Star, Pin Bars, Engulfing, Morning/Evening Star, Harami, Three White Soldiers and Three Black Crows
• Chart patterns: Head & Shoulders (and inverse), Double Top/Bottom, triangles, Cup & Handle, rising and falling wedges
• Structure: Support and resistance zones ranked by how often price touched them, plus order blocks and fair value gaps
• Market behavior panel: trend (slope + ADX), momentum (RSI + MACD), volatility (ATR), a Wyckoff-style market phase and an overall bias score

Good first extensions: add a new candlestick pattern to _candlePatterns.detect, add a Bollinger Band layer to _indicators, or add a new show... toggle in defaultParams.

Tips before you publish

• Return an array as long as data. Use null for candles that don't have a value yet.
• Guard against short history. Return early if there are too few candles for your longest period.
• Treat patterns as signals, not guarantees. The rules in NEXUS are simple heuristics, so tune sensitivity and validate any idea with backtesting before trading it live.
• Start on a demo account and watch how your indicator behaves on live ticks before you rely on it.

Challenges along the way

A few themes any platform like this has to deal with:

• Real-time data is unforgiving. Dropped connections, duplicate ticks and out-of-order messages all need handling.
• Backtests can lie. That's why I added walk-forward testing instead of trusting a single in-sample result.
• Risk controls can't be an afterthought. Limits and isolation belong in the core design, not bolted on.

What's next

• More strategies in the bot store
• Deeper backtesting tools
• A richer community and indicator-sharing layer
• Better docs and examples in the developer guide

If you have ideas, tell me in the comments.

Try it

• Platform: algocdk.site
• Developer Guide: algocdk.site/dev-guide
• Indicator Template: algocdk.site/indicator-template
• Bot Store: algocdk.site/botstore
• Live charts: algocdk.site/marketchart

I'd love feedback from fellow developers and traders. What would you want an algo trading platform to do that none of them do today?

---

Disclaimer: Trading derivatives and synthetic indices carries a high level of risk and you can lose your capital. Past performance does not guarantee future results. Nothing in this post is financial advice. Start on a demo account and only trade what you can afford to lose.